No Talent Shortages Here: The Real Challenges Facing Predictive Workforce Analytics
At APQC, we get a lot of questions about whether or not there is a shortage of data scientist talent. Recently, I had the opportunity to ask Greta Roberts, co-founder and CEO of Talent Analytics, Corp. for her perspective on this issue. Read her answer in this second blog post based on our interview. Greta’s answer surprised me and opened my eyes to a less talked about, but no less important, analytics talent shortage.
Elissa Tucker: There are a lot of business problems or use cases that predictive workforce analytics can help solve. However, doing so requires that there are enough data scientists to carry out the work. I want to get your perspective on this issue. Is there a data scientist talent shortage now or is one looming?
Greta Roberts: We actually do experience a significant shortage, but not of data scientists. We might be going against the grain a little bit, but we’ve never seen, never believed, or never experienced a shortage of data scientists. What we experience, and what we see repeatedly, is the shortage of managers or people that can identify a business problem that would be a great one for a data scientist to work on. So it’s actually the step before the data scientist where we see the shortage.
I’ll give you an example. We know one company—this is a real story—they hired 30 of the best data scientists they could possibly find. Then, they told them to go work on really important stuff. They said, “Assign great projects, do your magic, whatever it is.” After a year they looked at the data scientists’ productivity and saw there wasn’t really anything to show for this investment. They ended up firing all 30 of the great data scientists. It wasn’t the data scientists’ fault. This is really a true story. Business leaders are the ones that need to identify the projects for the data scientists to work on, and then let them run with it. The data scientists are not the people that should be finding the projects, themselves.
Imagine hiring great IT people, technology staff, and saying, go out and build us some great technology. And then the IT staff sit there and think: “Well, I don’t want to build anything because I don’t know what you need. I could spend all this money, and I could create these infrastructures and software applications. But I need to know what you guys need.” So the IT staff need to have the business leaders define what they need first, and then they build it.
One other way that we love to think about this at Talent Analytics, Corp., is in terms of supply and demand. First there is the demand. There is a real demand for something. Then you look for the supply, but demand always comes first. You don’t need a supply of data scientists until the managers create the demand for projects they have defined. We definitely, and in a very real way, see a shortage of defined projects, not the data scientists to do the work.
Elissa Tucker: What can business leaders do to end this shortage of defined analytics projects? How can they learn to ask business questions that predictive workforce analytics can help solve?
Greta Roberts: A lot of times business people or HR people think: “I’ve got to learn everything. I’ve got to learn how to define a problem. I’ve got to project manage it. I’ve got to do all the data science around it.” But there are really just two things that need to be prepared by a business or an HR person.
First, you need to learn what kinds of projects would make great workforce data science projects. If a predictive model is deployed, is it going to save money or make money for the firm? That has to be the number one reason for doing a project. Again, imagine IT. IT wouldn’t do a project unless they could say that the project would make the organization more efficient or save the organization money. There has to be a business reason for doing the project. It’s very easy to think: “Will we save money or will we make money on this project? Will this predictive model help us select more successful job candidates? Will it help us allocate assets to help plan retirement for people? Will it help create a succession plan or a pipeline where we can actually predict that there is going to be a better outcome?
Second, business leaders need to start by working on a project that uses, as much as possible, data that is already available. We have talked to a number of companies where they have said: “Yes, we can do this great predictive project, but we can’t do it for eight months because we have to, in the meantime, implement all this process where people start documenting all sorts of things so that we can begin gathering all kinds of data.” What happens is, you get this really negative feeling about even beginning the predictive project because people are thinking: “I have to add all of this administration on top of what we’re already doing.” So as much as possible use data that already exists. Then be smart. If there are a couple of little pieces here or there where you need to augment the data, just focus on those data points.
Other important considerations include, making sure you have enough data to find a credible, predictive pattern. We have people that will come to us and say: “We have 25 sales reps and we’d really like to understand what it is that makes them top performers.” There are only 25 people. It would be much preferred to go into an area that is rich with data. So you want to make sure that you give your data scientists enough data to play with because predictions are all about patterns.
Another thing that we see a lot of people do is go for the most glamorous analytics project that they can think of. For example, they think: “Let’s look at executives, and let’s predict something around executives.” But, there are not that many executives in even the largest organizations. We recommend going for less glamorous things, like the areas where the organization has huge volumes of employees such as call center representatives, sales representatives, bank tellers, or insurance agents. Look at areas where there is so much rich data. Saving just a little bit in those areas, or making just a little bit more money in those areas, is going to have a huge financial impact because there are so many people.
It is also important not to focus on an HR problem. I can’t tell you the number of failed predictive assignments that we’ve seen where HR people said: “We were given a mandate to work on predictive problems. We solved an HR problem, but it went nowhere because the problem didn’t matter to rest of the business.”
The last thing I would say to business leaders is make sure that when you’re predicting something, you predict what we call the final measure. For example, if I’m able to predict truck drivers who have a greater probability of having fewer accidents, that is amazing. People can immediately understand the ROI to the organization. This is not the case if I predict a middle measure or leading indicator such as predicting which employees will be highly engaged. That’s a middle measure and people rarely do the next step of saying: “If they’re engaged I can prove that our organizations save or we can make money due to certain engagement scores.”
Those would be the things that I would urge for the business leader and HR professional that is trying to identify a predictive analytics project. If they’re the data scientist, then there is a whole other list of things they need to know, like statistics, data mining, data preparation, data analysis, computer programming, and more. But, business leaders and HR professionals should not be doing the data scientist’s work. Their role is to be a scout, looking for projects that might be excellent predictive projects.
Want to know more? Greta will be a keynote speaker at The Predictive Analytics World for Workforce conference, which is being held in San Francisco from April 3rd through 6th in 2016. The conference is the premier workforce analytics event for HR professionals, business leaders, line of business managers, and analytics practitioners. This global, cross-industry event covers predictive solutions to today's greatest workforce challenges. Join Greta Roberts and APQC when you register today with 15% off code APQC15.
Listen to Elissa and Greta talk about getting started with predictive workforce analytics in the following webinar recording.Tweet